From CVRG Wiki
As an NHLBI resource, the CVRG is committed to the delivery of tools, deployed as analytic services on the grid, for analyzing cardiovascular data. Three different sets of tools are under development or have been deployed. These are:
- two ECG data analysis services
- heart shape and motion analysis services
- multi-scale biomarker discovery services
ECG Data Analysis Services
The CVRG ECG Gadget is an open source software platform developed within the CVRG Dashboard. A "gadget" is dynamic web content that can be placed on any web page. The CVRG ECG Gadget supports ECG time series data submission and management, visualization, annotation, and analysis. Currently two data analysis algorithms are deployed as services and are accessible from the ECG Gadget. Data may be passed into these algorithms, the results are retrieved by the gadget, a subset of analysis results is displayed, and users may download all numeric results and import them into an Excel spreadsheet.
These algorithms are:
Berger Algorithm Analytical Service
The first is a suite of algorithms developed by Dr. Ron Berger at Johns Hopkins University. See Berger, R.D., Kasper, E.K., Baughman K.L., Marban E., Calkins H., Tomaselli G.F. (1997) Beat-to-beat QT interval variability: novel evidence for repolarization lability in ischemic and nonischemic dilated cardiomyopathy. Circulation. 96(5):1557-1565 and Berger, R.D. (2003). QT Variability. J. Electrocardiol. 36: 83-87.
Physionet QT Algorithm Analytical (Chesnokov) Service
Heart Shape and Motion Analysis Services
The Large Deformation Diffeomorphic Metric Mapping (LDDMM) algorithm is the most widely used algorithm for analyzing anatomic shape and it’s variation Beg, M.F., Miller, M.I., Trouve, A., Younes, L. (2005). Computing Large Deformation Metric Mappings via Geodesic Flows of Diffeomorphisms. International Journal of Computer Vision 61, 139-157. It has been used extensively in brain research to discover shape-based biomarkers that are signatures of brain disease. We have recently used the LDDMM algorithm to study shape changes, assessed using MR and CT imagery collected from ex-vivo and in-vivo hearts, that are signatures of heart disease collected from ex-vivo and in-vivo hearts. [See Beg, M. F., Helm, P. A., McVeigh, E., Miller, M. I. and Winslow, R. L. (2004). Computational Cardiac Anatomy Using MRI. Magn. Reson. Med., 52(5): 1167 and Helm, P. et al (2006). Circ. Res. 98(1): 125-132. These algorithms are deployed through the LDDMM portlet, which in turn may be accessed on the CVRG Portal.
For more details on LDDMM, see the Project 4 wiki page on Grid-Tools for Cardiac Computational Anatomy.
The software application Computational Anatomy Works (CAWorks) was developed to support computational anatomy and shape analysis. The capabilities of CAWorks include:
- Interactive landmark placement to create segmentation (mask) of desired region of interest
- Specialized landmark placement plugins for subcortical structures such as hippocampus and amygdala
- Support for multiple Medical Imaging data formats, such as Nifti, Analyze, Freesurfer, DICOM and landmark data
- Quadra Planar view visualization
- Shape Analysis plugin modules, such as Large Deformation Diffeomorphic Metric Mapping (LDDMM):
More details on CAWorks can be found on the CVRG Website Here, and on the following CVRG wiki pages:
- CAWorks-template estimation
- CAWorks-cardiac remote processing
- CAWorks-ftp lddmm cardiac matching
- CAWorks-pipeline-canine heart
- CAWorks-pipeline-human heart
Multi-Scale Biomarker Discovery Services
Translational CV research more often than not involves the collection and analysis of what we refer to as multi-scale data sets (SNP, mRNA expression, protein expression, imaging, ECG, clinical, etc) for each patient in large cohorts. A challenge is to analyze such multi-scale data sets to discover biomarkers that are predictive of disease risk and treatment. Algorithms for biomarker discovery are available through the Biomarker discovery portlet. Algorithms include the K-TSP family of classifiers, known to be useful when the number of analysis features is large relative to the number of datasets. [See Geman D, d'Avignon C, Naiman DQ, Winslow RL. (2004) Classifying gene expression profiles from pairwise mRNA comparisons. Stat. Appl. Genet. Mol. Biol., 3(1): Article 19 and Xu L, Geman D, Winslow RL. (2007) Large-scale integration of cancer microarray data identifies a robust common cancer signature. BMC Bioinformatics. 8:275.
The algorithms developed by the CVRG team are all written in the R programming language, building upon various existing R packages. The CVRG team also developed an open source, web-based software platform for the storage, analysis, visualization and retrieval of data sets by the algorithms. More information on the software platform may be found here.
For more details on Statistical Learning with Multi-Scale Cardiovascular Data, visit the Project 5 section of the CVRG Wiki.